Optimization of the Multiechelon System for Repairable Spare Parts Using Swarm Intelligence Combined with a Local Search Strategy

摘要

Repairable spare parts are referred to critical and expensive components which could have failures; this type of spare parts are common in mining, military and other industries with high value physical assets. Repairable components, after a failure, can be restored to operation by a repair procedure which does not constitute an entire substitution. Many industries deploy their operations and their inventories in a geographically distributed structure. That distribution is composed by a central installation or depot and a set of bases where processes normally, take place (i.e. mining in northern Chile). This type of arrangement is called multi-echelon systems. The main concern in this type of systems is: in what number and how to distribute those expensive and critical repairable spare parts. The decision is restricted by a limited budget and the necessity of not affecting the normal operations of the system. Multi-echelon, multi-item optimization problems are known for their hardness in solving them to optimality, and therefore heuristics methods are approached to near-optimally solve such problems. The most prominent model is the Multi-Echelon Technique for Recoverable Item Control (METRIC), presented by Sherbrooke in 1968. That model has been extensively used in the military world and in the last years in others industries such aviation and mining. Through this model availability values are obtained from the performance characterization of backorders at the bases. METRIC allocates spare parts in the system on a global basis, since the METRIC model considers all locations simultaneously in the performance analysis. This work proposes the use of Particle Swarm Optimization with local search procedures to solve the multi-echelon of repairable spare parts optimization problem. The major difference between our proposal and previous works lies in that we will combine population based methods with specific local search methods The use of hybridization of non-traditional techniques to attain better optimization performance, is the main challenge of this work. No previous works have already been devoted to the use of hybridization of such techniques in such a type of problems.